Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -28620,6 +28620,72 @@
   return DAG.getNode(X86ISD::PSADBW, DL, SadVT, SadOp0, SadOp1);
 }
 
+// Attempt to replace an all_of/any_of style horizontal reduction with a MOVMSK.
+static SDValue combineHorizontalPredicateResult(SDNode *Extract,
+                                                SelectionDAG &DAG,
+                                                const X86Subtarget &Subtarget) {
+  if (!Subtarget.hasSSE2())
+    return SDValue();
+
+  EVT ExtractVT = Extract->getValueType(0);
+  unsigned BitWidth = ExtractVT.getSizeInBits();
+  if (ExtractVT != MVT::i64 && ExtractVT != MVT::i32 && ExtractVT != MVT::i16 &&
+      ExtractVT != MVT::i8)
+    return SDValue();
+
+  // Check for OR(any_of) and AND(all_of) horizontal reduction patterns.
+  for (ISD::NodeType Op : {ISD::OR, ISD::AND}) {
+    SDValue Match = matchBinOpReduction(Extract, Op);
+    if (!Match || (Match.getScalarValueSizeInBits() != BitWidth))
+      continue;
+
+    // We require AVX2 for PMOVMSKB for v16i16/v32i8;
+    unsigned MatchSizeInBits = Match.getValueSizeInBits();
+    if (!(MatchSizeInBits == 128 ||
+          (MatchSizeInBits == 256 &&
+           ((Subtarget.hasAVX() && BitWidth >= 32) || Subtarget.hasAVX2()))))
+      continue;
+
+    // Check that we are extracting a reduction of all sign bits.
+    if (DAG.ComputeNumSignBits(Match) != BitWidth)
+      continue;
+
+    // For 32/64 bit comparisons use MOVMSKPS/MOVMSKPD, else PMOVMSKB.
+    MVT MaskVT;
+    if (64 == BitWidth || 32 == BitWidth) {
+      MaskVT = MVT::getFloatingPointVT(BitWidth);
+      MaskVT = MVT::getVectorVT(MaskVT, MatchSizeInBits / BitWidth);
+    } else {
+      MaskVT = MVT::getVectorVT(MVT::i8, MatchSizeInBits / 8);
+    }
+
+    APInt CompareBits;
+    ISD::CondCode CondCode;
+    if (Op == ISD::OR) {
+      // any_of -> MOVMSK != 0
+      CompareBits = APInt::getNullValue(32);
+      CondCode = ISD::CondCode::SETNE;
+    } else {
+      // all_of -> MOVMSK == ((1 << NumElts) - 1)
+      CompareBits = APInt::getLowBitsSet(32, MaskVT.getVectorNumElements());
+      CondCode = ISD::CondCode::SETEQ;
+    }
+
+    SDLoc DL(Extract);
+    APInt ResOnes = APInt::getAllOnesValue(BitWidth);
+    APInt ResZero = APInt::getNullValue(BitWidth);
+
+    SDValue Res = DAG.getBitcast(MaskVT, Match);
+    Res = DAG.getNode(X86ISD::MOVMSK, DL, MVT::i32, Res);
+    Res = DAG.getSelectCC(DL, Res, DAG.getConstant(CompareBits, DL, MVT::i32),
+                          DAG.getConstant(ResOnes, DL, ExtractVT),
+                          DAG.getConstant(ResZero, DL, ExtractVT), CondCode);
+    return Res;
+  }
+
+  return SDValue();
+}
+
 static SDValue combineBasicSADPattern(SDNode *Extract, SelectionDAG &DAG,
                                       const X86Subtarget &Subtarget) {
   // PSADBW is only supported on SSE2 and up.
@@ -28731,6 +28797,9 @@
   if (SDValue SAD = combineBasicSADPattern(N, DAG, Subtarget))
     return SAD;
 
+  if (SDValue Cmp = combineHorizontalPredicateResult(N, DAG, Subtarget))
+    return Cmp;
+
   // Only operate on vectors of 4 elements, where the alternative shuffling
   // gets to be more expensive.
   if (InputVector.getValueType() != MVT::v4i32)
Index: test/CodeGen/X86/vector-compare-all_of.ll
===================================================================
--- test/CodeGen/X86/vector-compare-all_of.ll
+++ test/CodeGen/X86/vector-compare-all_of.ll
@@ -7,17 +7,21 @@
 ; SSE-LABEL: test_v2f64:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %rax
+; SSE-NEXT:    movmskpd %xmm1, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $3, %eax
+; SSE-NEXT:    movq $-1, %rax
+; SSE-NEXT:    cmovneq %rcx, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v2f64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vandpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovq %xmm0, %rax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $3, %eax
+; AVX-NEXT:    movq $-1, %rax
+; AVX-NEXT:    cmovneq %rcx, %rax
 ; AVX-NEXT:    retq
   %c = fcmp ogt <2 x double> %a0, %a1
   %s = sext <2 x i1> %c to <2 x i64>
@@ -33,32 +37,23 @@
 ; SSE-NEXT:    cmpltpd %xmm1, %xmm3
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm2
 ; SSE-NEXT:    andpd %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    movd %xmm0, %rax
+; SSE-NEXT:    movmskpd %xmm2, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $3, %eax
+; SSE-NEXT:    movq $-1, %rax
+; SSE-NEXT:    cmovneq %rcx, %rax
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: test_v4f64:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vandpd %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vandpd %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    vzeroupper
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: test_v4f64:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vandpd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; AVX-LABEL: test_v4f64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    vmovmskpd %ymm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $15, %eax
+; AVX-NEXT:    movq $-1, %rax
+; AVX-NEXT:    cmovneq %rcx, %rax
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
   %c = fcmp ogt <4 x double> %a0, %a1
   %s = sext <4 x i1> %c to <4 x i64>
   %1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
@@ -75,11 +70,11 @@
 ; SSE-NEXT:    cmpltpd %xmm1, %xmm3
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm2
 ; SSE-NEXT:    packsswb %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $15, %eax
+; SSE-NEXT:    movl $-1, %eax
+; SSE-NEXT:    cmovnel %ecx, %eax
 ; SSE-NEXT:    cltq
 ; SSE-NEXT:    retq
 ;
@@ -88,11 +83,11 @@
 ; AVX-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $15, %eax
+; AVX-NEXT:    movl $-1, %eax
+; AVX-NEXT:    cmovnel %ecx, %eax
 ; AVX-NEXT:    cltq
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -111,21 +106,21 @@
 ; SSE-LABEL: test_v4f32:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    cmpltps %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    movmskps %xmm1, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $15, %eax
+; SSE-NEXT:    movl $-1, %eax
+; SSE-NEXT:    cmovnel %ecx, %eax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v4f32:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vcmpltps %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vandpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $15, %eax
+; AVX-NEXT:    movl $-1, %eax
+; AVX-NEXT:    cmovnel %ecx, %eax
 ; AVX-NEXT:    retq
   %c = fcmp ogt <4 x float> %a0, %a1
   %s = sext <4 x i1> %c to <4 x i32>
@@ -143,38 +138,23 @@
 ; SSE-NEXT:    cmpltps %xmm1, %xmm3
 ; SSE-NEXT:    cmpltps %xmm0, %xmm2
 ; SSE-NEXT:    andps %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $15, %eax
+; SSE-NEXT:    movl $-1, %eax
+; SSE-NEXT:    cmovnel %ecx, %eax
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: test_v8f32:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    vzeroupper
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: test_v8f32:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; AVX-LABEL: test_v8f32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    vmovmskps %ymm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $255, %eax
+; AVX-NEXT:    movl $-1, %eax
+; AVX-NEXT:    cmovnel %ecx, %eax
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
   %c = fcmp ogt <8 x float> %a0, %a1
   %s = sext <8 x i1> %c to <8 x i32>
   %1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
@@ -193,14 +173,11 @@
 ; SSE-NEXT:    cmpltps %xmm1, %xmm3
 ; SSE-NEXT:    cmpltps %xmm0, %xmm2
 ; SSE-NEXT:    packsswb %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    pmovmskb %xmm2, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    movw $-1, %ax
+; SSE-NEXT:    cmovnew %cx, %ax
 ; SSE-NEXT:    cwtl
 ; SSE-NEXT:    retq
 ;
@@ -209,13 +186,11 @@
 ; AVX-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vpmovmskb %xmm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX-NEXT:    movw $-1, %ax
+; AVX-NEXT:    cmovnew %cx, %ax
 ; AVX-NEXT:    cwtl
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -236,17 +211,21 @@
 ; SSE-LABEL: test_v2i64:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtq %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %rax
+; SSE-NEXT:    movmskpd %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $3, %eax
+; SSE-NEXT:    movq $-1, %rax
+; SSE-NEXT:    cmovneq %rcx, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v2i64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovq %xmm0, %rax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $3, %eax
+; AVX-NEXT:    movq $-1, %rax
+; AVX-NEXT:    cmovneq %rcx, %rax
 ; AVX-NEXT:    retq
   %c = icmp sgt <2 x i64> %a0, %a1
   %s = sext <2 x i1> %c to <2 x i64>
@@ -262,9 +241,11 @@
 ; SSE-NEXT:    pcmpgtq %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtq %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %rax
+; SSE-NEXT:    movmskpd %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $3, %eax
+; SSE-NEXT:    movq $-1, %rax
+; SSE-NEXT:    cmovneq %rcx, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v4i64:
@@ -274,21 +255,22 @@
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovq %xmm0, %rax
+; AVX1-NEXT:    vmovmskpd %ymm0, %eax
+; AVX1-NEXT:    xorl %ecx, %ecx
+; AVX1-NEXT:    cmpl $15, %eax
+; AVX1-NEXT:    movq $-1, %rax
+; AVX1-NEXT:    cmovneq %rcx, %rax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v4i64:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovq %xmm0, %rax
+; AVX2-NEXT:    vmovmskpd %ymm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl $15, %eax
+; AVX2-NEXT:    movq $-1, %rax
+; AVX2-NEXT:    cmovneq %rcx, %rax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c = icmp sgt <4 x i64> %a0, %a1
@@ -307,11 +289,11 @@
 ; SSE-NEXT:    pcmpgtq %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtq %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $15, %eax
+; SSE-NEXT:    movl $-1, %eax
+; SSE-NEXT:    cmovnel %ecx, %eax
 ; SSE-NEXT:    cltq
 ; SSE-NEXT:    retq
 ;
@@ -322,11 +304,11 @@
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
+; AVX1-NEXT:    vmovmskps %xmm0, %eax
+; AVX1-NEXT:    xorl %ecx, %ecx
+; AVX1-NEXT:    cmpl $15, %eax
+; AVX1-NEXT:    movl $-1, %eax
+; AVX1-NEXT:    cmovnel %ecx, %eax
 ; AVX1-NEXT:    cltq
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
@@ -336,11 +318,11 @@
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
+; AVX2-NEXT:    vmovmskps %xmm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl $15, %eax
+; AVX2-NEXT:    movl $-1, %eax
+; AVX2-NEXT:    cmovnel %ecx, %eax
 ; AVX2-NEXT:    cltq
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -359,21 +341,21 @@
 ; SSE-LABEL: test_v4i32:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtd %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $15, %eax
+; SSE-NEXT:    movl $-1, %eax
+; SSE-NEXT:    cmovnel %ecx, %eax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v4i32:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $15, %eax
+; AVX-NEXT:    movl $-1, %eax
+; AVX-NEXT:    cmovnel %ecx, %eax
 ; AVX-NEXT:    retq
   %c = icmp sgt <4 x i32> %a0, %a1
   %s = sext <4 x i1> %c to <4 x i32>
@@ -391,11 +373,11 @@
 ; SSE-NEXT:    pcmpgtd %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtd %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $15, %eax
+; SSE-NEXT:    movl $-1, %eax
+; SSE-NEXT:    cmovnel %ecx, %eax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v8i32:
@@ -405,25 +387,22 @@
 ; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
+; AVX1-NEXT:    vmovmskps %ymm0, %eax
+; AVX1-NEXT:    xorl %ecx, %ecx
+; AVX1-NEXT:    cmpl $255, %eax
+; AVX1-NEXT:    movl $-1, %eax
+; AVX1-NEXT:    cmovnel %ecx, %eax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v8i32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl $255, %eax
+; AVX2-NEXT:    movl $-1, %eax
+; AVX2-NEXT:    cmovnel %ecx, %eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c = icmp sgt <8 x i32> %a0, %a1
@@ -444,14 +423,11 @@
 ; SSE-NEXT:    pcmpgtd %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtd %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    movw $-1, %ax
+; SSE-NEXT:    cmovnew %cx, %ax
 ; SSE-NEXT:    cwtl
 ; SSE-NEXT:    retq
 ;
@@ -462,13 +438,11 @@
 ; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
+; AVX1-NEXT:    vpmovmskb %xmm0, %eax
+; AVX1-NEXT:    xorl %ecx, %ecx
+; AVX1-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX1-NEXT:    movw $-1, %ax
+; AVX1-NEXT:    cmovnew %cx, %ax
 ; AVX1-NEXT:    cwtl
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
@@ -478,13 +452,11 @@
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX2-NEXT:    movw $-1, %ax
+; AVX2-NEXT:    cmovnew %cx, %ax
 ; AVX2-NEXT:    cwtl
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -505,28 +477,21 @@
 ; SSE-LABEL: test_v8i16:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
-; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    movw $-1, %ax
+; SSE-NEXT:    cmovnew %cx, %ax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v8i16:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
-; AVX-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX-NEXT:    vpmovmskb %xmm0, %eax
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX-NEXT:    movw $-1, %ax
+; AVX-NEXT:    cmovnew %cx, %ax
 ; AVX-NEXT:    retq
   %c = icmp sgt <8 x i16> %a0, %a1
   %s = sext <8 x i1> %c to <8 x i16>
@@ -546,15 +511,11 @@
 ; SSE-NEXT:    pcmpgtw %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtw %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
-; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    xorl %ecx, %ecx
+; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    movw $-1, %ax
+; SSE-NEXT:    cmovnew %cx, %ax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v16i16:
@@ -579,16 +540,11 @@
 ; AVX2-LABEL: test_v16i16:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl $-1, %eax
+; AVX2-NEXT:    movw $-1, %ax
+; AVX2-NEXT:    cmovnew %cx, %ax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c = icmp sgt <16 x i16> %a0, %a1
@@ -611,17 +567,13 @@
 ; SSE-NEXT:    pcmpgtw %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtw %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrlw $8, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    movb $-1, %al
+; SSE-NEXT:    je .LBB14_2
+; SSE-NEXT:  # BB#1:
+; SSE-NEXT:    xorl %eax, %eax
+; SSE-NEXT:  .LBB14_2:
 ; SSE-NEXT:    movsbl %al, %eax
 ; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
 ; SSE-NEXT:    retq
@@ -633,15 +585,13 @@
 ; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX1-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX1-NEXT:    movb $-1, %al
+; AVX1-NEXT:    cmpl $65535, %ecx # imm = 0xFFFF
+; AVX1-NEXT:    je .LBB14_2
+; AVX1-NEXT:  # BB#1:
+; AVX1-NEXT:    xorl %eax, %eax
+; AVX1-NEXT:  .LBB14_2:
 ; AVX1-NEXT:    movsbl %al, %eax
 ; AVX1-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX1-NEXT:    vzeroupper
@@ -652,15 +602,13 @@
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX2-NEXT:    movb $-1, %al
+; AVX2-NEXT:    cmpl $65535, %ecx # imm = 0xFFFF
+; AVX2-NEXT:    je .LBB14_2
+; AVX2-NEXT:  # BB#1:
+; AVX2-NEXT:    xorl %eax, %eax
+; AVX2-NEXT:  .LBB14_2:
 ; AVX2-NEXT:    movsbl %al, %eax
 ; AVX2-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
@@ -684,32 +632,26 @@
 ; SSE-LABEL: test_v16i8:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrlw $8, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    movb $-1, %al
+; SSE-NEXT:    je .LBB15_2
+; SSE-NEXT:  # BB#1:
+; SSE-NEXT:    xorl %eax, %eax
+; SSE-NEXT:  .LBB15_2:
 ; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v16i8:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX-NEXT:    movb $-1, %al
+; AVX-NEXT:    cmpl $65535, %ecx # imm = 0xFFFF
+; AVX-NEXT:    je .LBB15_2
+; AVX-NEXT:  # BB#1:
+; AVX-NEXT:    xorl %eax, %eax
+; AVX-NEXT:  .LBB15_2:
 ; AVX-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX-NEXT:    retq
   %c = icmp sgt <16 x i8> %a0, %a1
@@ -732,17 +674,13 @@
 ; SSE-NEXT:    pcmpgtb %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtb %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrlw $8, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    movb $-1, %al
+; SSE-NEXT:    je .LBB16_2
+; SSE-NEXT:  # BB#1:
+; SSE-NEXT:    xorl %eax, %eax
+; SSE-NEXT:  .LBB16_2:
 ; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE-NEXT:    retq
 ;
@@ -770,17 +708,13 @@
 ; AVX2-LABEL: test_v32i8:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtb %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %ecx
+; AVX2-NEXT:    movb $-1, %al
+; AVX2-NEXT:    cmpl $-1, %ecx
+; AVX2-NEXT:    je .LBB16_2
+; AVX2-NEXT:  # BB#1:
+; AVX2-NEXT:    xorl %eax, %eax
+; AVX2-NEXT:  .LBB16_2:
 ; AVX2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
Index: test/CodeGen/X86/vector-compare-any_of.ll
===================================================================
--- test/CodeGen/X86/vector-compare-any_of.ll
+++ test/CodeGen/X86/vector-compare-any_of.ll
@@ -7,17 +7,17 @@
 ; SSE-LABEL: test_v2f64:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %rax
+; SSE-NEXT:    movmskpd %xmm1, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbq %rax, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v2f64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vorpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovq %xmm0, %rax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbq %rax, %rax
 ; AVX-NEXT:    retq
   %c = fcmp ogt <2 x double> %a0, %a1
   %s = sext <2 x i1> %c to <2 x i64>
@@ -33,32 +33,19 @@
 ; SSE-NEXT:    cmpltpd %xmm1, %xmm3
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm2
 ; SSE-NEXT:    orpd %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movd %xmm0, %rax
+; SSE-NEXT:    movmskpd %xmm2, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbq %rax, %rax
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: test_v4f64:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vorpd %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vorpd %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    vzeroupper
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: test_v4f64:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vorpd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; AVX-LABEL: test_v4f64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    vmovmskpd %ymm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbq %rax, %rax
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
   %c = fcmp ogt <4 x double> %a0, %a1
   %s = sext <4 x i1> %c to <4 x i64>
   %1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
@@ -75,11 +62,9 @@
 ; SSE-NEXT:    cmpltpd %xmm1, %xmm3
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm2
 ; SSE-NEXT:    packsswb %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbl %eax, %eax
 ; SSE-NEXT:    cltq
 ; SSE-NEXT:    retq
 ;
@@ -88,11 +73,9 @@
 ; AVX-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbl %eax, %eax
 ; AVX-NEXT:    cltq
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -111,21 +94,17 @@
 ; SSE-LABEL: test_v4f32:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    cmpltps %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    movmskps %xmm1, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbl %eax, %eax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v4f32:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vcmpltps %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vorpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbl %eax, %eax
 ; AVX-NEXT:    retq
   %c = fcmp ogt <4 x float> %a0, %a1
   %s = sext <4 x i1> %c to <4 x i32>
@@ -143,38 +122,19 @@
 ; SSE-NEXT:    cmpltps %xmm1, %xmm3
 ; SSE-NEXT:    cmpltps %xmm0, %xmm2
 ; SSE-NEXT:    orps %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbl %eax, %eax
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: test_v8f32:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    vzeroupper
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: test_v8f32:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; AVX-LABEL: test_v8f32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    vmovmskps %ymm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbl %eax, %eax
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
   %c = fcmp ogt <8 x float> %a0, %a1
   %s = sext <8 x i1> %c to <8 x i32>
   %1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
@@ -193,14 +153,9 @@
 ; SSE-NEXT:    cmpltps %xmm1, %xmm3
 ; SSE-NEXT:    cmpltps %xmm0, %xmm2
 ; SSE-NEXT:    packsswb %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    pmovmskb %xmm2, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbw %ax, %ax
 ; SSE-NEXT:    cwtl
 ; SSE-NEXT:    retq
 ;
@@ -209,13 +164,9 @@
 ; AVX-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vpmovmskb %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbw %ax, %ax
 ; AVX-NEXT:    cwtl
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -236,17 +187,17 @@
 ; SSE-LABEL: test_v2i64:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtq %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %rax
+; SSE-NEXT:    movmskpd %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbq %rax, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v2i64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovq %xmm0, %rax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbq %rax, %rax
 ; AVX-NEXT:    retq
   %c = icmp sgt <2 x i64> %a0, %a1
   %s = sext <2 x i1> %c to <2 x i64>
@@ -262,9 +213,9 @@
 ; SSE-NEXT:    pcmpgtq %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtq %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %rax
+; SSE-NEXT:    movmskpd %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbq %rax, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v4i64:
@@ -274,21 +225,18 @@
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vorps %ymm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovq %xmm0, %rax
+; AVX1-NEXT:    vmovmskpd %ymm0, %eax
+; AVX1-NEXT:    negl %eax
+; AVX1-NEXT:    sbbq %rax, %rax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v4i64:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovq %xmm0, %rax
+; AVX2-NEXT:    vmovmskpd %ymm0, %eax
+; AVX2-NEXT:    negl %eax
+; AVX2-NEXT:    sbbq %rax, %rax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c = icmp sgt <4 x i64> %a0, %a1
@@ -307,11 +255,9 @@
 ; SSE-NEXT:    pcmpgtq %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtq %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbl %eax, %eax
 ; SSE-NEXT:    cltq
 ; SSE-NEXT:    retq
 ;
@@ -322,11 +268,9 @@
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
+; AVX1-NEXT:    vmovmskps %xmm0, %eax
+; AVX1-NEXT:    negl %eax
+; AVX1-NEXT:    sbbl %eax, %eax
 ; AVX1-NEXT:    cltq
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
@@ -336,11 +280,9 @@
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
+; AVX2-NEXT:    vmovmskps %xmm0, %eax
+; AVX2-NEXT:    negl %eax
+; AVX2-NEXT:    sbbl %eax, %eax
 ; AVX2-NEXT:    cltq
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -359,21 +301,17 @@
 ; SSE-LABEL: test_v4i32:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtd %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbl %eax, %eax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v4i32:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbl %eax, %eax
 ; AVX-NEXT:    retq
   %c = icmp sgt <4 x i32> %a0, %a1
   %s = sext <4 x i1> %c to <4 x i32>
@@ -391,11 +329,9 @@
 ; SSE-NEXT:    pcmpgtd %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtd %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movd %xmm0, %eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbl %eax, %eax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v8i32:
@@ -405,25 +341,18 @@
 ; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vorps %ymm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
+; AVX1-NEXT:    vmovmskps %ymm0, %eax
+; AVX1-NEXT:    negl %eax
+; AVX1-NEXT:    sbbl %eax, %eax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v8i32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    negl %eax
+; AVX2-NEXT:    sbbl %eax, %eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c = icmp sgt <8 x i32> %a0, %a1
@@ -444,14 +373,9 @@
 ; SSE-NEXT:    pcmpgtd %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtd %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbw %ax, %ax
 ; SSE-NEXT:    cwtl
 ; SSE-NEXT:    retq
 ;
@@ -462,13 +386,9 @@
 ; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovd %xmm0, %eax
+; AVX1-NEXT:    vpmovmskb %xmm0, %eax
+; AVX1-NEXT:    negl %eax
+; AVX1-NEXT:    sbbw %ax, %ax
 ; AVX1-NEXT:    cwtl
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
@@ -478,13 +398,9 @@
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    negl %eax
+; AVX2-NEXT:    sbbw %ax, %ax
 ; AVX2-NEXT:    cwtl
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -505,28 +421,17 @@
 ; SSE-LABEL: test_v8i16:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
-; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbw %ax, %ax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v8i16:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %xmm0, %eax
-; AVX-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX-NEXT:    vpmovmskb %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbw %ax, %ax
 ; AVX-NEXT:    retq
   %c = icmp sgt <8 x i16> %a0, %a1
   %s = sext <8 x i1> %c to <8 x i16>
@@ -546,15 +451,9 @@
 ; SSE-NEXT:    pcmpgtw %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtw %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movd %xmm1, %eax
-; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbw %ax, %ax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v16i16:
@@ -579,16 +478,9 @@
 ; AVX2-LABEL: test_v16i16:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; AVX2-NEXT:    negl %eax
+; AVX2-NEXT:    sbbw %ax, %ax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c = icmp sgt <16 x i16> %a0, %a1
@@ -611,19 +503,9 @@
 ; SSE-NEXT:    pcmpgtw %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtw %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrlw $8, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    movsbl %al, %eax
-; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbw %ax, %ax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v16i16_legal:
@@ -633,17 +515,9 @@
 ; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX1-NEXT:    movsbl %al, %eax
-; AVX1-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX1-NEXT:    vpmovmskb %xmm0, %eax
+; AVX1-NEXT:    negl %eax
+; AVX1-NEXT:    sbbw %ax, %ax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
@@ -652,17 +526,9 @@
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    movsbl %al, %eax
-; AVX2-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    negl %eax
+; AVX2-NEXT:    sbbw %ax, %ax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c  = icmp sgt <16 x i16> %a0, %a1
@@ -684,33 +550,17 @@
 ; SSE-LABEL: test_v16i8:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pcmpgtb %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrlw $8, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbb %al, %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v16i8:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX-NEXT:    vpmovmskb %xmm0, %eax
+; AVX-NEXT:    negl %eax
+; AVX-NEXT:    sbbb %al, %al
 ; AVX-NEXT:    retq
   %c = icmp sgt <16 x i8> %a0, %a1
   %s = sext <16 x i1> %c to <16 x i8>
@@ -732,18 +582,9 @@
 ; SSE-NEXT:    pcmpgtb %xmm3, %xmm1
 ; SSE-NEXT:    pcmpgtb %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrlw $8, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    sbbb %al, %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v32i8:
@@ -770,18 +611,9 @@
 ; AVX2-LABEL: test_v32i8:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpgtb %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; AVX2-NEXT:    negl %eax
+; AVX2-NEXT:    sbbb %al, %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %c  = icmp sgt <32 x i8> %a0, %a1